Robot Learning with Super-Linear Scaling

TL;DR

This paper introduces CASHER, a scalable robot learning pipeline that leverages crowdsourcing and simulation to reduce human effort, enabling efficient policy training and adaptation across diverse real-world tasks.

Contribution

CASHER is a novel pipeline that combines crowdsourcing, digital twin creation, and simulation-based data collection to achieve super-linear scaling in robot learning with minimal human effort.

Findings

CASHER achieves zero-shot and few-shot scaling on real-world tasks.

The pipeline enables fine-tuning policies using only video scans.

Performance improves with reduced human effort over training progress.

Abstract

Scaling robot learning requires data collection pipelines that scale favorably with human effort. In this work, we propose Crowdsourcing and Amortizing Human Effort for Real-to-Sim-to-Real(CASHER), a pipeline for scaling up data collection and learning in simulation where the performance scales superlinearly with human effort. The key idea is to crowdsource digital twins of real-world scenes using 3D reconstruction and collect large-scale data in simulation, rather than the real-world. Data collection in simulation is initially driven by RL, bootstrapped with human demonstrations. As the training of a generalist policy progresses across environments, its generalization capabilities can be used to replace human effort with model generated demonstrations. This results in a pipeline where behavioral data is collected in simulation with continually reducing human effort. We show that CASHER demonstrates zero-shot and few-shot scaling laws on three real-world tasks across diverse scenarios. We show that CASHER enables fine-tuning of pre-trained policies to a target scenario using a video scan without any additional human effort. See our project website: https://casher-robot-learning.github.io/CASHER/